一种机器人智能语音识别算法研究

A Voice Recognition Algorithm for Intelligent Robot

摘要:

针对智能机器人在非特定人语音识别中识别率偏低的问题,提出了一种双门限的端点检测算法,精确地检测出了语音端点,对分形维数和Mel频率倒谱系数(MFCC)进行结合,同时基于隐马尔可夫(HMM)模型,提出了智能机器人命令识别系统；在实验室环境下,利用Cool Edit软件录制了5男5女的语音,采样率为8 kHz,精度为16位,内容为5个命令词,每个词均被采集6次,将每人的前3次发音作为模板语音,后3次发音作为测试语音,实验结果表明,系统识别率可以达到85%以上,MFCC与分形维数混合的语音特征参数的算法提高了系统识别率,优化了系统性能；该方法用于非特定人语音智能识别是可行的、有效的。

Abstract:

For problem of the low speaker-independent voice recognition rate in intelligent robots, a double threshold endpoint detection algorithm is proposed, which is able to accurately detect voice endpoints, and fractal dimension and Mel Frequency Cepstrum Coefficient (MFCC) is combined, at the same time, based on Hidden Markov Model (HMM), an intelligent robot command recognition system is proposed. In the laboratory environment, Cool Edit software is used to record voice of five men and five woman, sampling rate is 8kHz, accuracy is 16,5 command words, each word is collected 6 times, The first three pronunciation is seen as a template for voice, the rest is seen as a test voice, The experimental results show that System identification rate can reach more than 85%,and the voice feature parameters of the algorithm of MFCC mixed with fractal dimension can improve the system recognition rate and optimizes the system performance. The method used for non-specific voice intelligent recognition is feasible and effective.